You need to sign in or sign up before continuing.
write.c 40.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
/*
 * linux/fs/nfs/write.c
 *
 * Writing file data over NFS.
 *
 * We do it like this: When a (user) process wishes to write data to an
 * NFS file, a write request is allocated that contains the RPC task data
 * plus some info on the page to be written, and added to the inode's
 * write chain. If the process writes past the end of the page, an async
 * RPC call to write the page is scheduled immediately; otherwise, the call
 * is delayed for a few seconds.
 *
 * Just like readahead, no async I/O is performed if wsize < PAGE_SIZE.
 *
 * Write requests are kept on the inode's writeback list. Each entry in
 * that list references the page (portion) to be written. When the
 * cache timeout has expired, the RPC task is woken up, and tries to
 * lock the page. As soon as it manages to do so, the request is moved
 * from the writeback list to the writelock list.
 *
 * Note: we must make sure never to confuse the inode passed in the
 * write_page request with the one in page->inode. As far as I understand
 * it, these are different when doing a swap-out.
 *
 * To understand everything that goes on here and in the NFS read code,
 * one should be aware that a page is locked in exactly one of the following
 * cases:
 *
 *  -	A write request is in progress.
 *  -	A user process is in generic_file_write/nfs_update_page
 *  -	A user process is in generic_file_read
 *
 * Also note that because of the way pages are invalidated in
 * nfs_revalidate_inode, the following assertions hold:
 *
 *  -	If a page is dirty, there will be no read requests (a page will
 *	not be re-read unless invalidated by nfs_revalidate_inode).
 *  -	If the page is not uptodate, there will be no pending write
 *	requests, and no process will be in nfs_update_page.
 *
 * FIXME: Interaction with the vmscan routines is not optimal yet.
 * Either vmscan must be made nfs-savvy, or we need a different page
 * reclaim concept that supports something like FS-independent
 * buffer_heads with a b_ops-> field.
 *
 * Copyright (C) 1996, 1997, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/mpage.h>
#include <linux/writeback.h>

#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs_page.h>
#include <asm/uaccess.h>
#include <linux/smp_lock.h>

#include "delegation.h"
C
Chuck Lever 已提交
65
#include "iostat.h"
L
Linus Torvalds 已提交
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82

#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

#define MIN_POOL_WRITE		(32)
#define MIN_POOL_COMMIT		(4)

/*
 * Local function declarations
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context*,
					    struct inode *,
					    struct page *,
					    unsigned int, unsigned int);
static int nfs_wait_on_write_congestion(struct address_space *, int);
static int nfs_wait_on_requests(struct inode *, unsigned long, unsigned int);
static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
			   unsigned int npages, int how);
83 84 85
static const struct rpc_call_ops nfs_write_partial_ops;
static const struct rpc_call_ops nfs_write_full_ops;
static const struct rpc_call_ops nfs_commit_ops;
L
Linus Torvalds 已提交
86 87

static kmem_cache_t *nfs_wdata_cachep;
88
static mempool_t *nfs_wdata_mempool;
L
Linus Torvalds 已提交
89 90 91 92
static mempool_t *nfs_commit_mempool;

static DECLARE_WAIT_QUEUE_HEAD(nfs_write_congestion);

93
struct nfs_write_data *nfs_commit_alloc(void)
L
Linus Torvalds 已提交
94 95
{
	struct nfs_write_data *p = mempool_alloc(nfs_commit_mempool, SLAB_NOFS);
96

L
Linus Torvalds 已提交
97 98 99 100 101 102 103
	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
	}
	return p;
}

104
void nfs_commit_free(struct nfs_write_data *p)
L
Linus Torvalds 已提交
105
{
106 107
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
L
Linus Torvalds 已提交
108 109 110
	mempool_free(p, nfs_commit_mempool);
}

111
struct nfs_write_data *nfs_writedata_alloc(size_t len)
112
{
113
	unsigned int pagecount = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
114 115 116 117 118
	struct nfs_write_data *p = mempool_alloc(nfs_wdata_mempool, SLAB_NOFS);

	if (p) {
		memset(p, 0, sizeof(*p));
		INIT_LIST_HEAD(&p->pages);
119
		p->npages = pagecount;
120 121
		if (pagecount <= ARRAY_SIZE(p->page_array))
			p->pagevec = p->page_array;
122
		else {
123 124
			p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS);
			if (!p->pagevec) {
125 126 127 128 129 130 131 132
				mempool_free(p, nfs_wdata_mempool);
				p = NULL;
			}
		}
	}
	return p;
}

A
Adrian Bunk 已提交
133
static void nfs_writedata_free(struct nfs_write_data *p)
134 135 136 137 138 139
{
	if (p && (p->pagevec != &p->page_array[0]))
		kfree(p->pagevec);
	mempool_free(p, nfs_wdata_mempool);
}

140
void nfs_writedata_release(void *wdata)
L
Linus Torvalds 已提交
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
{
	nfs_writedata_free(wdata);
}

/* Adjust the file length if we're writing beyond the end */
static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int count)
{
	struct inode *inode = page->mapping->host;
	loff_t end, i_size = i_size_read(inode);
	unsigned long end_index = (i_size - 1) >> PAGE_CACHE_SHIFT;

	if (i_size > 0 && page->index < end_index)
		return;
	end = ((loff_t)page->index << PAGE_CACHE_SHIFT) + ((loff_t)offset+count);
	if (i_size >= end)
		return;
C
Chuck Lever 已提交
157
	nfs_inc_stats(inode, NFSIOS_EXTENDWRITE);
L
Linus Torvalds 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
	i_size_write(inode, end);
}

/* We can set the PG_uptodate flag if we see that a write request
 * covers the full page.
 */
static void nfs_mark_uptodate(struct page *page, unsigned int base, unsigned int count)
{
	loff_t end_offs;

	if (PageUptodate(page))
		return;
	if (base != 0)
		return;
	if (count == PAGE_CACHE_SIZE) {
		SetPageUptodate(page);
		return;
	}

	end_offs = i_size_read(page->mapping->host) - 1;
	if (end_offs < 0)
		return;
	/* Is this the last page? */
	if (page->index != (unsigned long)(end_offs >> PAGE_CACHE_SHIFT))
		return;
	/* This is the last page: set PG_uptodate if we cover the entire
	 * extent of the data, then zero the rest of the page.
	 */
	if (count == (unsigned int)(end_offs & (PAGE_CACHE_SIZE - 1)) + 1) {
		memclear_highpage_flush(page, count, PAGE_CACHE_SIZE - count);
		SetPageUptodate(page);
	}
}

/*
 * Write a page synchronously.
 * Offset is the data offset within the page.
 */
static int nfs_writepage_sync(struct nfs_open_context *ctx, struct inode *inode,
		struct page *page, unsigned int offset, unsigned int count,
		int how)
{
	unsigned int	wsize = NFS_SERVER(inode)->wsize;
	int		result, written = 0;
	struct nfs_write_data *wdata;

204
	wdata = nfs_writedata_alloc(wsize);
L
Linus Torvalds 已提交
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
	if (!wdata)
		return -ENOMEM;

	wdata->flags = how;
	wdata->cred = ctx->cred;
	wdata->inode = inode;
	wdata->args.fh = NFS_FH(inode);
	wdata->args.context = ctx;
	wdata->args.pages = &page;
	wdata->args.stable = NFS_FILE_SYNC;
	wdata->args.pgbase = offset;
	wdata->args.count = wsize;
	wdata->res.fattr = &wdata->fattr;
	wdata->res.verf = &wdata->verf;

	dprintk("NFS:      nfs_writepage_sync(%s/%Ld %d@%Ld)\n",
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count, (long long)(page_offset(page) + offset));

225
	set_page_writeback(page);
L
Linus Torvalds 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
	nfs_begin_data_update(inode);
	do {
		if (count < wsize)
			wdata->args.count = count;
		wdata->args.offset = page_offset(page) + wdata->args.pgbase;

		result = NFS_PROTO(inode)->write(wdata);

		if (result < 0) {
			/* Must mark the page invalid after I/O error */
			ClearPageUptodate(page);
			goto io_error;
		}
		if (result < wdata->args.count)
			printk(KERN_WARNING "NFS: short write, count=%u, result=%d\n",
					wdata->args.count, result);

		wdata->args.offset += result;
	        wdata->args.pgbase += result;
		written += result;
		count -= result;
C
Chuck Lever 已提交
247
		nfs_add_stats(inode, NFSIOS_SERVERWRITTENBYTES, result);
L
Linus Torvalds 已提交
248 249 250 251 252 253 254 255 256 257
	} while (count);
	/* Update file length */
	nfs_grow_file(page, offset, written);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, written);

	if (PageError(page))
		ClearPageError(page);

io_error:
258
	nfs_end_data_update(inode);
259
	end_page_writeback(page);
L
Linus Torvalds 已提交
260 261 262 263 264 265 266 267 268 269 270
	nfs_writedata_free(wdata);
	return written ? written : result;
}

static int nfs_writepage_async(struct nfs_open_context *ctx,
		struct inode *inode, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_page	*req;

	req = nfs_update_request(ctx, inode, page, offset, count);
271 272
	if (IS_ERR(req))
		return PTR_ERR(req);
L
Linus Torvalds 已提交
273 274 275 276 277
	/* Update file length */
	nfs_grow_file(page, offset, count);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, offset, count);
	nfs_unlock_request(req);
278
	return 0;
L
Linus Torvalds 已提交
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
}

static int wb_priority(struct writeback_control *wbc)
{
	if (wbc->for_reclaim)
		return FLUSH_HIGHPRI;
	if (wbc->for_kupdate)
		return FLUSH_LOWPRI;
	return 0;
}

/*
 * Write an mmapped page to the server.
 */
int nfs_writepage(struct page *page, struct writeback_control *wbc)
{
	struct nfs_open_context *ctx;
	struct inode *inode = page->mapping->host;
	unsigned long end_index;
	unsigned offset = PAGE_CACHE_SIZE;
	loff_t i_size = i_size_read(inode);
	int inode_referenced = 0;
	int priority = wb_priority(wbc);
	int err;

C
Chuck Lever 已提交
304 305 306
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, 1);

L
Linus Torvalds 已提交
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332
	/*
	 * Note: We need to ensure that we have a reference to the inode
	 *       if we are to do asynchronous writes. If not, waiting
	 *       in nfs_wait_on_request() may deadlock with clear_inode().
	 *
	 *       If igrab() fails here, then it is in any case safe to
	 *       call nfs_wb_page(), since there will be no pending writes.
	 */
	if (igrab(inode) != 0)
		inode_referenced = 1;
	end_index = i_size >> PAGE_CACHE_SHIFT;

	/* Ensure we've flushed out any previous writes */
	nfs_wb_page_priority(inode, page, priority);

	/* easy case */
	if (page->index < end_index)
		goto do_it;
	/* things got complicated... */
	offset = i_size & (PAGE_CACHE_SIZE-1);

	/* OK, are we completely out? */
	err = 0; /* potential race with truncate - ignore */
	if (page->index >= end_index+1 || !offset)
		goto out;
do_it:
333
	ctx = nfs_find_open_context(inode, NULL, FMODE_WRITE);
L
Linus Torvalds 已提交
334 335 336 337 338 339 340
	if (ctx == NULL) {
		err = -EBADF;
		goto out;
	}
	lock_kernel();
	if (!IS_SYNC(inode) && inode_referenced) {
		err = nfs_writepage_async(ctx, inode, page, 0, offset);
341 342
		if (!wbc->for_writepages)
			nfs_flush_inode(inode, 0, 0, wb_priority(wbc));
L
Linus Torvalds 已提交
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370
	} else {
		err = nfs_writepage_sync(ctx, inode, page, 0,
						offset, priority);
		if (err >= 0) {
			if (err != offset)
				redirty_page_for_writepage(wbc, page);
			err = 0;
		}
	}
	unlock_kernel();
	put_nfs_open_context(ctx);
out:
	unlock_page(page);
	if (inode_referenced)
		iput(inode);
	return err; 
}

/*
 * Note: causes nfs_update_request() to block on the assumption
 * 	 that the writeback is generated due to memory pressure.
 */
int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc)
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	struct inode *inode = mapping->host;
	int err;

C
Chuck Lever 已提交
371 372
	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGES);

L
Linus Torvalds 已提交
373 374 375 376 377 378 379 380 381 382 383
	err = generic_writepages(mapping, wbc);
	if (err)
		return err;
	while (test_and_set_bit(BDI_write_congested, &bdi->state) != 0) {
		if (wbc->nonblocking)
			return 0;
		nfs_wait_on_write_congestion(mapping, 0);
	}
	err = nfs_flush_inode(inode, 0, 0, wb_priority(wbc));
	if (err < 0)
		goto out;
C
Chuck Lever 已提交
384
	nfs_add_stats(inode, NFSIOS_WRITEPAGES, err);
L
Linus Torvalds 已提交
385 386 387 388 389 390
	wbc->nr_to_write -= err;
	if (!wbc->nonblocking && wbc->sync_mode == WB_SYNC_ALL) {
		err = nfs_wait_on_requests(inode, 0, 0);
		if (err < 0)
			goto out;
	}
391
	err = nfs_commit_inode(inode, wb_priority(wbc));
L
Linus Torvalds 已提交
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
	if (err > 0) {
		wbc->nr_to_write -= err;
		err = 0;
	}
out:
	clear_bit(BDI_write_congested, &bdi->state);
	wake_up_all(&nfs_write_congestion);
	return err;
}

/*
 * Insert a write request into an inode
 */
static int nfs_inode_add_request(struct inode *inode, struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	int error;

	error = radix_tree_insert(&nfsi->nfs_page_tree, req->wb_index, req);
	BUG_ON(error == -EEXIST);
	if (error)
		return error;
	if (!nfsi->npages) {
		igrab(inode);
		nfs_begin_data_update(inode);
		if (nfs_have_delegation(inode, FMODE_WRITE))
			nfsi->change_attr++;
	}
420
	SetPagePrivate(req->wb_page);
L
Linus Torvalds 已提交
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
	nfsi->npages++;
	atomic_inc(&req->wb_count);
	return 0;
}

/*
 * Insert a write request into an inode
 */
static void nfs_inode_remove_request(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	BUG_ON (!NFS_WBACK_BUSY(req));

	spin_lock(&nfsi->req_lock);
437
	ClearPagePrivate(req->wb_page);
L
Linus Torvalds 已提交
438 439 440 441
	radix_tree_delete(&nfsi->nfs_page_tree, req->wb_index);
	nfsi->npages--;
	if (!nfsi->npages) {
		spin_unlock(&nfsi->req_lock);
442
		nfs_end_data_update(inode);
L
Linus Torvalds 已提交
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
		iput(inode);
	} else
		spin_unlock(&nfsi->req_lock);
	nfs_clear_request(req);
	nfs_release_request(req);
}

/*
 * Find a request
 */
static inline struct nfs_page *
_nfs_find_request(struct inode *inode, unsigned long index)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page *req;

	req = (struct nfs_page*)radix_tree_lookup(&nfsi->nfs_page_tree, index);
	if (req)
		atomic_inc(&req->wb_count);
	return req;
}

static struct nfs_page *
nfs_find_request(struct inode *inode, unsigned long index)
{
	struct nfs_page		*req;
	struct nfs_inode	*nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
	req = _nfs_find_request(inode, index);
	spin_unlock(&nfsi->req_lock);
	return req;
}

/*
 * Add a request to the inode's dirty list.
 */
static void
nfs_mark_request_dirty(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
487 488
	radix_tree_tag_set(&nfsi->nfs_page_tree,
			req->wb_index, NFS_PAGE_TAG_DIRTY);
L
Linus Torvalds 已提交
489 490 491
	nfs_list_add_request(req, &nfsi->dirty);
	nfsi->ndirty++;
	spin_unlock(&nfsi->req_lock);
492
	inc_zone_page_state(req->wb_page, NR_FILE_DIRTY);
L
Linus Torvalds 已提交
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
	mark_inode_dirty(inode);
}

/*
 * Check if a request is dirty
 */
static inline int
nfs_dirty_request(struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
	return !list_empty(&req->wb_list) && req->wb_list_head == &nfsi->dirty;
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
/*
 * Add a request to the inode's commit list.
 */
static void
nfs_mark_request_commit(struct nfs_page *req)
{
	struct inode *inode = req->wb_context->dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&nfsi->req_lock);
	nfs_list_add_request(req, &nfsi->commit);
	nfsi->ncommit++;
	spin_unlock(&nfsi->req_lock);
520
	inc_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
L
Linus Torvalds 已提交
521 522 523 524 525 526 527 528 529
	mark_inode_dirty(inode);
}
#endif

/*
 * Wait for a request to complete.
 *
 * Interruptible by signals only if mounted with intr flag.
 */
530
static int nfs_wait_on_requests_locked(struct inode *inode, unsigned long idx_start, unsigned int npages)
L
Linus Torvalds 已提交
531 532 533 534 535 536 537 538 539 540 541 542 543
{
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page *req;
	unsigned long		idx_end, next;
	unsigned int		res = 0;
	int			error;

	if (npages == 0)
		idx_end = ~0;
	else
		idx_end = idx_start + npages - 1;

	next = idx_start;
544
	while (radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree, (void **)&req, next, 1, NFS_PAGE_TAG_WRITEBACK)) {
L
Linus Torvalds 已提交
545 546 547 548
		if (req->wb_index > idx_end)
			break;

		next = req->wb_index + 1;
549
		BUG_ON(!NFS_WBACK_BUSY(req));
L
Linus Torvalds 已提交
550 551 552 553 554

		atomic_inc(&req->wb_count);
		spin_unlock(&nfsi->req_lock);
		error = nfs_wait_on_request(req);
		nfs_release_request(req);
555
		spin_lock(&nfsi->req_lock);
L
Linus Torvalds 已提交
556 557 558 559 560 561 562
		if (error < 0)
			return error;
		res++;
	}
	return res;
}

563 564 565 566 567 568 569 570 571 572 573
static int nfs_wait_on_requests(struct inode *inode, unsigned long idx_start, unsigned int npages)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	int ret;

	spin_lock(&nfsi->req_lock);
	ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
	spin_unlock(&nfsi->req_lock);
	return ret;
}

T
Trond Myklebust 已提交
574
static void nfs_cancel_dirty_list(struct list_head *head)
575 576 577 578 579 580 581 582 583 584
{
	struct nfs_page *req;
	while(!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_inode_remove_request(req);
		nfs_clear_page_writeback(req);
	}
}

T
Trond Myklebust 已提交
585 586 587 588 589 590 591 592 593
static void nfs_cancel_commit_list(struct list_head *head)
{
	struct nfs_page *req;

	while(!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_inode_remove_request(req);
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
594
		nfs_clear_page_writeback(req);
T
Trond Myklebust 已提交
595 596 597
	}
}

L
Linus Torvalds 已提交
598 599 600 601 602 603 604 605 606 607 608 609 610 611
/*
 * nfs_scan_dirty - Scan an inode for dirty requests
 * @inode: NFS inode to scan
 * @dst: destination list
 * @idx_start: lower bound of page->index to scan.
 * @npages: idx_start + npages sets the upper bound to scan.
 *
 * Moves requests from the inode's dirty page list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
static int
nfs_scan_dirty(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
{
	struct nfs_inode *nfsi = NFS_I(inode);
612 613 614 615 616 617 618 619
	int res = 0;

	if (nfsi->ndirty != 0) {
		res = nfs_scan_lock_dirty(nfsi, dst, idx_start, npages);
		nfsi->ndirty -= res;
		if ((nfsi->ndirty == 0) != list_empty(&nfsi->dirty))
			printk(KERN_ERR "NFS: desynchronized value of nfs_i.ndirty.\n");
	}
L
Linus Torvalds 已提交
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
	return res;
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
/*
 * nfs_scan_commit - Scan an inode for commit requests
 * @inode: NFS inode to scan
 * @dst: destination list
 * @idx_start: lower bound of page->index to scan.
 * @npages: idx_start + npages sets the upper bound to scan.
 *
 * Moves requests from the inode's 'commit' request list.
 * The requests are *not* checked to ensure that they form a contiguous set.
 */
static int
nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
{
	struct nfs_inode *nfsi = NFS_I(inode);
638 639 640
	int res = 0;

	if (nfsi->ncommit != 0) {
641
		res = nfs_scan_list(nfsi, &nfsi->commit, dst, idx_start, npages);
642 643 644 645
		nfsi->ncommit -= res;
		if ((nfsi->ncommit == 0) != list_empty(&nfsi->commit))
			printk(KERN_ERR "NFS: desynchronized value of nfs_i.ncommit.\n");
	}
L
Linus Torvalds 已提交
646 647
	return res;
}
648 649 650 651 652
#else
static inline int nfs_scan_commit(struct inode *inode, struct list_head *dst, unsigned long idx_start, unsigned int npages)
{
	return 0;
}
L
Linus Torvalds 已提交
653 654 655 656 657 658 659 660 661 662 663 664
#endif

static int nfs_wait_on_write_congestion(struct address_space *mapping, int intr)
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	DEFINE_WAIT(wait);
	int ret = 0;

	might_sleep();

	if (!bdi_write_congested(bdi))
		return 0;
C
Chuck Lever 已提交
665 666 667

	nfs_inc_stats(mapping->host, NFSIOS_CONGESTIONWAIT);

L
Linus Torvalds 已提交
668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722
	if (intr) {
		struct rpc_clnt *clnt = NFS_CLIENT(mapping->host);
		sigset_t oldset;

		rpc_clnt_sigmask(clnt, &oldset);
		prepare_to_wait(&nfs_write_congestion, &wait, TASK_INTERRUPTIBLE);
		if (bdi_write_congested(bdi)) {
			if (signalled())
				ret = -ERESTARTSYS;
			else
				schedule();
		}
		rpc_clnt_sigunmask(clnt, &oldset);
	} else {
		prepare_to_wait(&nfs_write_congestion, &wait, TASK_UNINTERRUPTIBLE);
		if (bdi_write_congested(bdi))
			schedule();
	}
	finish_wait(&nfs_write_congestion, &wait);
	return ret;
}


/*
 * Try to update any existing write request, or create one if there is none.
 * In order to match, the request's credentials must match those of
 * the calling process.
 *
 * Note: Should always be called with the Page Lock held!
 */
static struct nfs_page * nfs_update_request(struct nfs_open_context* ctx,
		struct inode *inode, struct page *page,
		unsigned int offset, unsigned int bytes)
{
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs_page		*req, *new = NULL;
	unsigned long		rqend, end;

	end = offset + bytes;

	if (nfs_wait_on_write_congestion(page->mapping, server->flags & NFS_MOUNT_INTR))
		return ERR_PTR(-ERESTARTSYS);
	for (;;) {
		/* Loop over all inode entries and see if we find
		 * A request for the page we wish to update
		 */
		spin_lock(&nfsi->req_lock);
		req = _nfs_find_request(inode, page->index);
		if (req) {
			if (!nfs_lock_request_dontget(req)) {
				int error;
				spin_unlock(&nfsi->req_lock);
				error = nfs_wait_on_request(req);
				nfs_release_request(req);
N
Neil Brown 已提交
723 724 725
				if (error < 0) {
					if (new)
						nfs_release_request(new);
L
Linus Torvalds 已提交
726
					return ERR_PTR(error);
N
Neil Brown 已提交
727
				}
L
Linus Torvalds 已提交
728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
				continue;
			}
			spin_unlock(&nfsi->req_lock);
			if (new)
				nfs_release_request(new);
			break;
		}

		if (new) {
			int error;
			nfs_lock_request_dontget(new);
			error = nfs_inode_add_request(inode, new);
			if (error) {
				spin_unlock(&nfsi->req_lock);
				nfs_unlock_request(new);
				return ERR_PTR(error);
			}
			spin_unlock(&nfsi->req_lock);
			nfs_mark_request_dirty(new);
			return new;
		}
		spin_unlock(&nfsi->req_lock);

		new = nfs_create_request(ctx, inode, page, offset, bytes);
		if (IS_ERR(new))
			return new;
	}

	/* We have a request for our page.
	 * If the creds don't match, or the
	 * page addresses don't match,
	 * tell the caller to wait on the conflicting
	 * request.
	 */
	rqend = req->wb_offset + req->wb_bytes;
	if (req->wb_context != ctx
	    || req->wb_page != page
	    || !nfs_dirty_request(req)
	    || offset > rqend || end < req->wb_offset) {
		nfs_unlock_request(req);
		return ERR_PTR(-EBUSY);
	}

	/* Okay, the request matches. Update the region */
	if (offset < req->wb_offset) {
		req->wb_offset = offset;
		req->wb_pgbase = offset;
		req->wb_bytes = rqend - req->wb_offset;
	}

	if (end > rqend)
		req->wb_bytes = end - req->wb_offset;

	return req;
}

int nfs_flush_incompatible(struct file *file, struct page *page)
{
	struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
	struct inode	*inode = page->mapping->host;
	struct nfs_page	*req;
	int		status = 0;
	/*
	 * Look for a request corresponding to this page. If there
	 * is one, and it belongs to another file, we flush it out
	 * before we try to copy anything into the page. Do this
	 * due to the lack of an ACCESS-type call in NFSv2.
	 * Also do the same if we find a request from an existing
	 * dropped page.
	 */
	req = nfs_find_request(inode, page->index);
	if (req) {
		if (req->wb_page != page || ctx != req->wb_context)
			status = nfs_wb_page(inode, page);
		nfs_release_request(req);
	}
	return (status < 0) ? status : 0;
}

/*
 * Update and possibly write a cached page of an NFS file.
 *
 * XXX: Keep an eye on generic_file_read to make sure it doesn't do bad
 * things with a page scheduled for an RPC call (e.g. invalidate it).
 */
int nfs_updatepage(struct file *file, struct page *page,
		unsigned int offset, unsigned int count)
{
	struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
	struct inode	*inode = page->mapping->host;
	struct nfs_page	*req;
	int		status = 0;

C
Chuck Lever 已提交
821 822
	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);

L
Linus Torvalds 已提交
823
	dprintk("NFS:      nfs_updatepage(%s/%s %d@%Ld)\n",
824 825 826
		file->f_dentry->d_parent->d_name.name,
		file->f_dentry->d_name.name, count,
		(long long)(page_offset(page) +offset));
L
Linus Torvalds 已提交
827 828 829 830 831 832 833 834 835 836 837 838 839 840 841

	if (IS_SYNC(inode)) {
		status = nfs_writepage_sync(ctx, inode, page, offset, count, 0);
		if (status > 0) {
			if (offset == 0 && status == PAGE_CACHE_SIZE)
				SetPageUptodate(page);
			return 0;
		}
		return status;
	}

	/* If we're not using byte range locks, and we know the page
	 * is entirely in cache, it may be more efficient to avoid
	 * fragmenting write requests.
	 */
842
	if (PageUptodate(page) && inode->i_flock == NULL && !(file->f_mode & O_SYNC)) {
L
Linus Torvalds 已提交
843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
		loff_t end_offs = i_size_read(inode) - 1;
		unsigned long end_index = end_offs >> PAGE_CACHE_SHIFT;

		count += offset;
		offset = 0;
		if (unlikely(end_offs < 0)) {
			/* Do nothing */
		} else if (page->index == end_index) {
			unsigned int pglen;
			pglen = (unsigned int)(end_offs & (PAGE_CACHE_SIZE-1)) + 1;
			if (count < pglen)
				count = pglen;
		} else if (page->index < end_index)
			count = PAGE_CACHE_SIZE;
	}

	/*
	 * Try to find an NFS request corresponding to this page
	 * and update it.
	 * If the existing request cannot be updated, we must flush
	 * it out now.
	 */
	do {
		req = nfs_update_request(ctx, inode, page, offset, count);
		status = (IS_ERR(req)) ? PTR_ERR(req) : 0;
		if (status != -EBUSY)
			break;
		/* Request could not be updated. Flush it out and try again */
		status = nfs_wb_page(inode, page);
	} while (status >= 0);
	if (status < 0)
		goto done;

	status = 0;

	/* Update file length */
	nfs_grow_file(page, offset, count);
	/* Set the PG_uptodate flag? */
	nfs_mark_uptodate(page, req->wb_pgbase, req->wb_bytes);
	nfs_unlock_request(req);
done:
        dprintk("NFS:      nfs_updatepage returns %d (isize %Ld)\n",
			status, (long long)i_size_read(inode));
	if (status < 0)
		ClearPageUptodate(page);
	return status;
}

static void nfs_writepage_release(struct nfs_page *req)
{
	end_page_writeback(req->wb_page);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (!PageError(req->wb_page)) {
		if (NFS_NEED_RESCHED(req)) {
			nfs_mark_request_dirty(req);
			goto out;
		} else if (NFS_NEED_COMMIT(req)) {
			nfs_mark_request_commit(req);
			goto out;
		}
	}
	nfs_inode_remove_request(req);

out:
	nfs_clear_commit(req);
	nfs_clear_reschedule(req);
#else
	nfs_inode_remove_request(req);
#endif
913
	nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
}

static inline int flush_task_priority(int how)
{
	switch (how & (FLUSH_HIGHPRI|FLUSH_LOWPRI)) {
		case FLUSH_HIGHPRI:
			return RPC_PRIORITY_HIGH;
		case FLUSH_LOWPRI:
			return RPC_PRIORITY_LOW;
	}
	return RPC_PRIORITY_NORMAL;
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_write_rpcsetup(struct nfs_page *req,
		struct nfs_write_data *data,
932
		const struct rpc_call_ops *call_ops,
L
Linus Torvalds 已提交
933 934 935 936
		unsigned int count, unsigned int offset,
		int how)
{
	struct inode		*inode;
937
	int flags;
L
Linus Torvalds 已提交
938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955

	/* Set up the RPC argument and reply structs
	 * NB: take care not to mess about with data->commit et al. */

	data->req = req;
	data->inode = inode = req->wb_context->dentry->d_inode;
	data->cred = req->wb_context->cred;

	data->args.fh     = NFS_FH(inode);
	data->args.offset = req_offset(req) + offset;
	data->args.pgbase = req->wb_pgbase + offset;
	data->args.pages  = data->pagevec;
	data->args.count  = count;
	data->args.context = req->wb_context;

	data->res.fattr   = &data->fattr;
	data->res.count   = count;
	data->res.verf    = &data->verf;
956
	nfs_fattr_init(&data->fattr);
L
Linus Torvalds 已提交
957

958 959 960
	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
L
Linus Torvalds 已提交
961 962 963 964 965 966
	NFS_PROTO(inode)->write_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;

	dprintk("NFS: %4d initiated write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
967
		data->task.tk_pid,
L
Linus Torvalds 已提交
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
		inode->i_sb->s_id,
		(long long)NFS_FILEID(inode),
		count,
		(unsigned long long)data->args.offset);
}

static void nfs_execute_write(struct nfs_write_data *data)
{
	struct rpc_clnt *clnt = NFS_CLIENT(data->inode);
	sigset_t oldset;

	rpc_clnt_sigmask(clnt, &oldset);
	lock_kernel();
	rpc_execute(&data->task);
	unlock_kernel();
	rpc_clnt_sigunmask(clnt, &oldset);
}

/*
 * Generate multiple small requests to write out a single
 * contiguous dirty area on one page.
 */
T
Trond Myklebust 已提交
990
static int nfs_flush_multi(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
991 992 993 994
{
	struct nfs_page *req = nfs_list_entry(head->next);
	struct page *page = req->wb_page;
	struct nfs_write_data *data;
995 996
	size_t wsize = NFS_SERVER(inode)->wsize, nbytes;
	unsigned int offset;
L
Linus Torvalds 已提交
997 998 999 1000 1001 1002
	int requests = 0;
	LIST_HEAD(list);

	nfs_list_remove_request(req);

	nbytes = req->wb_bytes;
1003 1004 1005 1006
	do {
		size_t len = min(nbytes, wsize);

		data = nfs_writedata_alloc(len);
L
Linus Torvalds 已提交
1007 1008 1009 1010
		if (!data)
			goto out_bad;
		list_add(&data->pages, &list);
		requests++;
1011 1012
		nbytes -= len;
	} while (nbytes != 0);
L
Linus Torvalds 已提交
1013 1014 1015
	atomic_set(&req->wb_complete, requests);

	ClearPageError(page);
1016
	set_page_writeback(page);
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021 1022 1023 1024 1025
	offset = 0;
	nbytes = req->wb_bytes;
	do {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del_init(&data->pages);

		data->pagevec[0] = page;

		if (nbytes > wsize) {
1026 1027
			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
					wsize, offset, how);
L
Linus Torvalds 已提交
1028 1029 1030
			offset += wsize;
			nbytes -= wsize;
		} else {
1031 1032
			nfs_write_rpcsetup(req, data, &nfs_write_partial_ops,
					nbytes, offset, how);
L
Linus Torvalds 已提交
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
			nbytes = 0;
		}
		nfs_execute_write(data);
	} while (nbytes != 0);

	return 0;

out_bad:
	while (!list_empty(&list)) {
		data = list_entry(list.next, struct nfs_write_data, pages);
		list_del(&data->pages);
		nfs_writedata_free(data);
	}
	nfs_mark_request_dirty(req);
1047
	nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
	return -ENOMEM;
}

/*
 * Create an RPC task for the given write request and kick it.
 * The page must have been locked by the caller.
 *
 * It may happen that the page we're passed is not marked dirty.
 * This is the case if nfs_updatepage detects a conflicting request
 * that has been written but not committed.
 */
T
Trond Myklebust 已提交
1059
static int nfs_flush_one(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064 1065
{
	struct nfs_page		*req;
	struct page		**pages;
	struct nfs_write_data	*data;
	unsigned int		count;

1066
	data = nfs_writedata_alloc(NFS_SERVER(inode)->wsize);
L
Linus Torvalds 已提交
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
	if (!data)
		goto out_bad;

	pages = data->pagevec;
	count = 0;
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &data->pages);
		ClearPageError(req->wb_page);
1077
		set_page_writeback(req->wb_page);
L
Linus Torvalds 已提交
1078 1079 1080 1081 1082 1083
		*pages++ = req->wb_page;
		count += req->wb_bytes;
	}
	req = nfs_list_entry(data->pages.next);

	/* Set up the argument struct */
1084
	nfs_write_rpcsetup(req, data, &nfs_write_full_ops, count, 0, how);
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089 1090 1091 1092

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		struct nfs_page *req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
1093
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1094 1095 1096 1097
	}
	return -ENOMEM;
}

T
Trond Myklebust 已提交
1098
static int nfs_flush_list(struct inode *inode, struct list_head *head, int npages, int how)
L
Linus Torvalds 已提交
1099 1100
{
	LIST_HEAD(one_request);
T
Trond Myklebust 已提交
1101 1102 1103 1104 1105
	int (*flush_one)(struct inode *, struct list_head *, int);
	struct nfs_page	*req;
	int wpages = NFS_SERVER(inode)->wpages;
	int wsize = NFS_SERVER(inode)->wsize;
	int error;
L
Linus Torvalds 已提交
1106

T
Trond Myklebust 已提交
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
	flush_one = nfs_flush_one;
	if (wsize < PAGE_CACHE_SIZE)
		flush_one = nfs_flush_multi;
	/* For single writes, FLUSH_STABLE is more efficient */
	if (npages <= wpages && npages == NFS_I(inode)->npages
			&& nfs_list_entry(head->next)->wb_bytes <= wsize)
		how |= FLUSH_STABLE;

	do {
		nfs_coalesce_requests(head, &one_request, wpages);
L
Linus Torvalds 已提交
1117
		req = nfs_list_entry(one_request.next);
T
Trond Myklebust 已提交
1118
		error = flush_one(inode, &one_request, how);
L
Linus Torvalds 已提交
1119
		if (error < 0)
T
Trond Myklebust 已提交
1120 1121 1122 1123
			goto out_err;
	} while (!list_empty(head));
	return 0;
out_err:
L
Linus Torvalds 已提交
1124 1125 1126 1127
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_dirty(req);
1128
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1129 1130 1131 1132 1133 1134 1135
	}
	return error;
}

/*
 * Handle a write reply that flushed part of a page.
 */
1136
static void nfs_writeback_done_partial(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1137
{
1138
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1139 1140 1141 1142 1143 1144 1145 1146 1147
	struct nfs_page		*req = data->req;
	struct page		*page = req->wb_page;

	dprintk("NFS: write (%s/%Ld %d@%Ld)",
		req->wb_context->dentry->d_inode->i_sb->s_id,
		(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
		req->wb_bytes,
		(long long)req_offset(req));

1148 1149 1150 1151
	if (nfs_writeback_done(task, data) != 0)
		return;

	if (task->tk_status < 0) {
L
Linus Torvalds 已提交
1152 1153
		ClearPageUptodate(page);
		SetPageError(page);
1154 1155
		req->wb_context->error = task->tk_status;
		dprintk(", error = %d\n", task->tk_status);
L
Linus Torvalds 已提交
1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
	} else {
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->verf.committed < NFS_FILE_SYNC) {
			if (!NFS_NEED_COMMIT(req)) {
				nfs_defer_commit(req);
				memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
				dprintk(" defer commit\n");
			} else if (memcmp(&req->wb_verf, &data->verf, sizeof(req->wb_verf))) {
				nfs_defer_reschedule(req);
				dprintk(" server reboot detected\n");
			}
		} else
#endif
			dprintk(" OK\n");
	}

	if (atomic_dec_and_test(&req->wb_complete))
		nfs_writepage_release(req);
}

1176 1177 1178 1179 1180
static const struct rpc_call_ops nfs_write_partial_ops = {
	.rpc_call_done = nfs_writeback_done_partial,
	.rpc_release = nfs_writedata_release,
};

L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186 1187
/*
 * Handle a write reply that flushes a whole page.
 *
 * FIXME: There is an inherent race with invalidate_inode_pages and
 *	  writebacks since the page->count is kept > 1 for as long
 *	  as the page has a write request pending.
 */
1188
static void nfs_writeback_done_full(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1189
{
1190
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1191 1192 1193
	struct nfs_page		*req;
	struct page		*page;

1194 1195 1196
	if (nfs_writeback_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
	/* Update attributes as result of writeback. */
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
		page = req->wb_page;

		dprintk("NFS: write (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));

1209
		if (task->tk_status < 0) {
L
Linus Torvalds 已提交
1210 1211
			ClearPageUptodate(page);
			SetPageError(page);
1212
			req->wb_context->error = task->tk_status;
L
Linus Torvalds 已提交
1213 1214
			end_page_writeback(page);
			nfs_inode_remove_request(req);
1215
			dprintk(", error = %d\n", task->tk_status);
L
Linus Torvalds 已提交
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
			goto next;
		}
		end_page_writeback(page);

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
		if (data->args.stable != NFS_UNSTABLE || data->verf.committed == NFS_FILE_SYNC) {
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		memcpy(&req->wb_verf, &data->verf, sizeof(req->wb_verf));
		nfs_mark_request_commit(req);
		dprintk(" marked for commit\n");
#else
		nfs_inode_remove_request(req);
#endif
	next:
1233
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1234 1235 1236
	}
}

1237 1238 1239 1240 1241 1242
static const struct rpc_call_ops nfs_write_full_ops = {
	.rpc_call_done = nfs_writeback_done_full,
	.rpc_release = nfs_writedata_release,
};


L
Linus Torvalds 已提交
1243 1244 1245
/*
 * This function is called when the WRITE call is complete.
 */
1246
int nfs_writeback_done(struct rpc_task *task, struct nfs_write_data *data)
L
Linus Torvalds 已提交
1247 1248 1249
{
	struct nfs_writeargs	*argp = &data->args;
	struct nfs_writeres	*resp = &data->res;
1250
	int status;
L
Linus Torvalds 已提交
1251 1252 1253 1254

	dprintk("NFS: %4d nfs_writeback_done (status %d)\n",
		task->tk_pid, task->tk_status);

1255 1256 1257 1258
	/* Call the NFS version-specific code */
	status = NFS_PROTO(data->inode)->write_done(task, data);
	if (status != 0)
		return status;
C
Chuck Lever 已提交
1259 1260
	nfs_add_stats(data->inode, NFSIOS_SERVERWRITTENBYTES, resp->count);

L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
	if (resp->verf->committed < argp->stable && task->tk_status >= 0) {
		/* We tried a write call, but the server did not
		 * commit data to stable storage even though we
		 * requested it.
		 * Note: There is a known bug in Tru64 < 5.0 in which
		 *	 the server reports NFS_DATA_SYNC, but performs
		 *	 NFS_FILE_SYNC. We therefore implement this checking
		 *	 as a dprintk() in order to avoid filling syslog.
		 */
		static unsigned long    complain;

		if (time_before(complain, jiffies)) {
			dprintk("NFS: faulty NFS server %s:"
				" (committed = %d) != (stable = %d)\n",
				NFS_SERVER(data->inode)->hostname,
				resp->verf->committed, argp->stable);
			complain = jiffies + 300 * HZ;
		}
	}
#endif
	/* Is this a short write? */
	if (task->tk_status >= 0 && resp->count < argp->count) {
		static unsigned long    complain;

C
Chuck Lever 已提交
1286 1287
		nfs_inc_stats(data->inode, NFSIOS_SHORTWRITE);

L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
		/* Has the server at least made some progress? */
		if (resp->count != 0) {
			/* Was this an NFSv2 write or an NFSv3 stable write? */
			if (resp->verf->committed != NFS_UNSTABLE) {
				/* Resend from where the server left off */
				argp->offset += resp->count;
				argp->pgbase += resp->count;
				argp->count -= resp->count;
			} else {
				/* Resend as a stable write in order to avoid
				 * headaches in the case of a server crash.
				 */
				argp->stable = NFS_FILE_SYNC;
			}
			rpc_restart_call(task);
1303
			return -EAGAIN;
L
Linus Torvalds 已提交
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
		}
		if (time_before(complain, jiffies)) {
			printk(KERN_WARNING
			       "NFS: Server wrote zero bytes, expected %u.\n",
					argp->count);
			complain = jiffies + 300 * HZ;
		}
		/* Can't do anything about it except throw an error. */
		task->tk_status = -EIO;
	}
1314
	return 0;
L
Linus Torvalds 已提交
1315 1316 1317 1318
}


#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
1319
void nfs_commit_release(void *wdata)
L
Linus Torvalds 已提交
1320 1321 1322 1323 1324 1325 1326 1327
{
	nfs_commit_free(wdata);
}

/*
 * Set up the argument/result storage required for the RPC call.
 */
static void nfs_commit_rpcsetup(struct list_head *head,
1328 1329
		struct nfs_write_data *data,
		int how)
L
Linus Torvalds 已提交
1330
{
1331
	struct nfs_page		*first;
L
Linus Torvalds 已提交
1332
	struct inode		*inode;
1333
	int flags;
L
Linus Torvalds 已提交
1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345

	/* Set up the RPC argument and reply structs
	 * NB: take care not to mess about with data->commit et al. */

	list_splice_init(head, &data->pages);
	first = nfs_list_entry(data->pages.next);
	inode = first->wb_context->dentry->d_inode;

	data->inode	  = inode;
	data->cred	  = first->wb_context->cred;

	data->args.fh     = NFS_FH(data->inode);
1346 1347 1348 1349
	/* Note: we always request a commit of the entire inode */
	data->args.offset = 0;
	data->args.count  = 0;
	data->res.count   = 0;
L
Linus Torvalds 已提交
1350 1351
	data->res.fattr   = &data->fattr;
	data->res.verf    = &data->verf;
1352
	nfs_fattr_init(&data->fattr);
1353 1354 1355 1356

	/* Set up the initial task struct.  */
	flags = (how & FLUSH_SYNC) ? 0 : RPC_TASK_ASYNC;
	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, &nfs_commit_ops, data);
L
Linus Torvalds 已提交
1357 1358 1359 1360 1361
	NFS_PROTO(inode)->commit_setup(data, how);

	data->task.tk_priority = flush_task_priority(how);
	data->task.tk_cookie = (unsigned long)inode;
	
1362
	dprintk("NFS: %4d initiated commit call\n", data->task.tk_pid);
L
Linus Torvalds 已提交
1363 1364 1365 1366 1367 1368
}

/*
 * Commit dirty pages
 */
static int
1369
nfs_commit_list(struct inode *inode, struct list_head *head, int how)
L
Linus Torvalds 已提交
1370 1371 1372 1373
{
	struct nfs_write_data	*data;
	struct nfs_page         *req;

1374
	data = nfs_commit_alloc();
L
Linus Torvalds 已提交
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388

	if (!data)
		goto out_bad;

	/* Set up the argument struct */
	nfs_commit_rpcsetup(head, data, how);

	nfs_execute_write(data);
	return 0;
 out_bad:
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_mark_request_commit(req);
T
Trond Myklebust 已提交
1389
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
1390
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1391 1392 1393 1394 1395 1396 1397
	}
	return -ENOMEM;
}

/*
 * COMMIT call returned
 */
1398
static void nfs_commit_done(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
1399
{
1400
	struct nfs_write_data	*data = calldata;
L
Linus Torvalds 已提交
1401 1402 1403 1404 1405
	struct nfs_page		*req;

        dprintk("NFS: %4d nfs_commit_done (status %d)\n",
                                task->tk_pid, task->tk_status);

1406 1407 1408 1409
	/* Call the NFS version-specific code */
	if (NFS_PROTO(data->inode)->commit_done(task, data) != 0)
		return;

L
Linus Torvalds 已提交
1410 1411 1412
	while (!list_empty(&data->pages)) {
		req = nfs_list_entry(data->pages.next);
		nfs_list_remove_request(req);
1413
		dec_zone_page_state(req->wb_page, NR_UNSTABLE_NFS);
L
Linus Torvalds 已提交
1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438

		dprintk("NFS: commit (%s/%Ld %d@%Ld)",
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
			req->wb_bytes,
			(long long)req_offset(req));
		if (task->tk_status < 0) {
			req->wb_context->error = task->tk_status;
			nfs_inode_remove_request(req);
			dprintk(", error = %d\n", task->tk_status);
			goto next;
		}

		/* Okay, COMMIT succeeded, apparently. Check the verifier
		 * returned by the server against all stored verfs. */
		if (!memcmp(req->wb_verf.verifier, data->verf.verifier, sizeof(data->verf.verifier))) {
			/* We have a match */
			nfs_inode_remove_request(req);
			dprintk(" OK\n");
			goto next;
		}
		/* We have a mismatch. Write the page again */
		dprintk(" mismatch\n");
		nfs_mark_request_dirty(req);
	next:
1439
		nfs_clear_page_writeback(req);
L
Linus Torvalds 已提交
1440 1441
	}
}
1442 1443 1444 1445 1446

static const struct rpc_call_ops nfs_commit_ops = {
	.rpc_call_done = nfs_commit_done,
	.rpc_release = nfs_commit_release,
};
1447 1448 1449 1450 1451
#else
static inline int nfs_commit_list(struct inode *inode, struct list_head *head, int how)
{
	return 0;
}
L
Linus Torvalds 已提交
1452 1453 1454 1455 1456 1457 1458
#endif

static int nfs_flush_inode(struct inode *inode, unsigned long idx_start,
			   unsigned int npages, int how)
{
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
T
Trond Myklebust 已提交
1459
	int res;
L
Linus Torvalds 已提交
1460 1461 1462 1463

	spin_lock(&nfsi->req_lock);
	res = nfs_scan_dirty(inode, &head, idx_start, npages);
	spin_unlock(&nfsi->req_lock);
1464
	if (res) {
T
Trond Myklebust 已提交
1465 1466 1467
		int error = nfs_flush_list(inode, &head, res, how);
		if (error < 0)
			return error;
1468
	}
L
Linus Torvalds 已提交
1469 1470 1471 1472
	return res;
}

#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
1473
int nfs_commit_inode(struct inode *inode, int how)
L
Linus Torvalds 已提交
1474 1475 1476
{
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
T
Trond Myklebust 已提交
1477
	int res;
L
Linus Torvalds 已提交
1478 1479

	spin_lock(&nfsi->req_lock);
1480 1481
	res = nfs_scan_commit(inode, &head, 0, 0);
	spin_unlock(&nfsi->req_lock);
L
Linus Torvalds 已提交
1482
	if (res) {
T
Trond Myklebust 已提交
1483
		int error = nfs_commit_list(inode, &head, how);
1484 1485 1486
		if (error < 0)
			return error;
	}
L
Linus Torvalds 已提交
1487 1488 1489 1490
	return res;
}
#endif

1491 1492
int nfs_sync_inode_wait(struct inode *inode, unsigned long idx_start,
		unsigned int npages, int how)
L
Linus Torvalds 已提交
1493
{
1494 1495
	struct nfs_inode *nfsi = NFS_I(inode);
	LIST_HEAD(head);
1496
	int nocommit = how & FLUSH_NOCOMMIT;
1497
	int pages, ret;
L
Linus Torvalds 已提交
1498

1499 1500
	how &= ~FLUSH_NOCOMMIT;
	spin_lock(&nfsi->req_lock);
L
Linus Torvalds 已提交
1501
	do {
1502 1503
		ret = nfs_wait_on_requests_locked(inode, idx_start, npages);
		if (ret != 0)
1504
			continue;
1505 1506 1507
		pages = nfs_scan_dirty(inode, &head, idx_start, npages);
		if (pages != 0) {
			spin_unlock(&nfsi->req_lock);
1508
			if (how & FLUSH_INVALIDATE)
T
Trond Myklebust 已提交
1509
				nfs_cancel_dirty_list(&head);
1510 1511
			else
				ret = nfs_flush_list(inode, &head, pages, how);
1512 1513 1514 1515 1516
			spin_lock(&nfsi->req_lock);
			continue;
		}
		if (nocommit)
			break;
1517
		pages = nfs_scan_commit(inode, &head, idx_start, npages);
1518 1519
		if (pages == 0)
			break;
1520 1521
		if (how & FLUSH_INVALIDATE) {
			spin_unlock(&nfsi->req_lock);
T
Trond Myklebust 已提交
1522
			nfs_cancel_commit_list(&head);
1523 1524 1525 1526
			spin_lock(&nfsi->req_lock);
			continue;
		}
		pages += nfs_scan_commit(inode, &head, 0, 0);
1527 1528 1529 1530 1531 1532
		spin_unlock(&nfsi->req_lock);
		ret = nfs_commit_list(inode, &head, how);
		spin_lock(&nfsi->req_lock);
	} while (ret >= 0);
	spin_unlock(&nfsi->req_lock);
	return ret;
L
Linus Torvalds 已提交
1533 1534
}

D
David Howells 已提交
1535
int __init nfs_init_writepagecache(void)
L
Linus Torvalds 已提交
1536 1537 1538 1539 1540 1541 1542 1543
{
	nfs_wdata_cachep = kmem_cache_create("nfs_write_data",
					     sizeof(struct nfs_write_data),
					     0, SLAB_HWCACHE_ALIGN,
					     NULL, NULL);
	if (nfs_wdata_cachep == NULL)
		return -ENOMEM;

1544 1545
	nfs_wdata_mempool = mempool_create_slab_pool(MIN_POOL_WRITE,
						     nfs_wdata_cachep);
L
Linus Torvalds 已提交
1546 1547 1548
	if (nfs_wdata_mempool == NULL)
		return -ENOMEM;

1549 1550
	nfs_commit_mempool = mempool_create_slab_pool(MIN_POOL_COMMIT,
						      nfs_wdata_cachep);
L
Linus Torvalds 已提交
1551 1552 1553 1554 1555 1556
	if (nfs_commit_mempool == NULL)
		return -ENOMEM;

	return 0;
}

1557
void nfs_destroy_writepagecache(void)
L
Linus Torvalds 已提交
1558 1559 1560 1561 1562 1563 1564
{
	mempool_destroy(nfs_commit_mempool);
	mempool_destroy(nfs_wdata_mempool);
	if (kmem_cache_destroy(nfs_wdata_cachep))
		printk(KERN_INFO "nfs_write_data: not all structures were freed\n");
}